HR: 1340h
AN: IN43B-0336 [Abstracts]
TI: VLab: A Grid for Planetary Materials Computations
AU: * da Silva, C R
EM: cesards@msi.umn.edu
AF: University of Minnesota, Minnesota Supercomputing Institute
117 Pleasant St. SE, Minneapolis, MN 55414
United States
AU: Erlebacher, G
EM: erlebach@csit.fsu.edu
AF: Florida State University, 489 Dirac Science Library, Tallahassee, FL 32306-4120
United States
AU: Pierce, M
EM: mpierce@cs.indiana.edu
AF: Indiana University, Community Grids Lab
501 N. Morton St., Bloomington, IN 47404
United States
AU: Karki, B
EM: karki@bit.csc.lsu.edu
AF: Louisiana State University, Department of Computer Science
283 Coates Hall, Baton Rouge, LA 70803
United States
AU: Yuen, D A
EM: davey@krissy.geo.umn.edu
AF: University of Minnesota, Dept. of Geology and Geophysics
23 Pillsbury Hall, Minneapolis, MN 55455
United States
AU: Wentzcovitch, R M
EM: wentzcov@cems.umn.edu
AF: University of Minnesota, Dept. Chem. Eng. Mat. Sci.
421 Washington Avenue SE, Minneapolis, MN 55414
United States
AB:
Computational studies of planetary materials have grown tremendously in
complexity in the last decade. A typical first principles high PT
elasticity calculation is a time consuming procedure involving hundreds to
thousands of job submissions and runs. These numerous runs result from
parameter space sampling (phonons, pressures, strains, composition, etc)
following a well defined workflow. However, most runs can be decoupled
with little or no overhead and scheduled for concurrent execution on
distributed systems. In addition, human intervention can be minimized by
automating workflow executions.
We describe a dedicated grid to prepare, submit, monitor, and analyze
runs' results. The back-end is a collection of web-services providing
access to distributed computing nodes, controlling workflow execution,
monitoring services, and providing data analyses tools, visualization
services, data bases, and authentication services. The front-end consists
of second generation portals interfacing users with the underlying grid
and providing a consistent abstraction of the entire collection of
services as a single system. The presentation layer makes extensive use of
portlets. They can be tailored to fit the required task granularity on
both front- and back-end sides, and to provide a well structured user
interface. Narada-Brokering is used as the middleware to provide
abstractions for locations and availability of back-end services and
persistence/fault tolerance.
UR: http://www.vlab.msi.umn.edu/
DE: 0525 Data management
DE: 0545 Modeling (4255)
DE: 0560 Numerical solutions (4255)
DE: 0599 General or miscellaneous
SC: Earth and Space Science Informatics [IN]
MN: Fall Meeting 2005